A Novel Distance Estimation Framework for PDR Based Indoor Localization Using RNNs

Next-generation communication technologies such as millimeter-wave wireless local area networks (mmWave WLANs) use multiple antenna-based directional beamforming where the location of the communicating entities plays a crucial role. Currently, the standards such as IEEE 802.11ad and 11ay employ beamforming training for the stations (STAs) and access points (APs) to estimate the sector or direction in which the communication is feasible and to establish the link successfully. However, frequent beam training increases power consumption and reduces channel utilization. Considering the location of AP is fixed, the frequent beam training of the STAs can be limited if the STAs can estimate the location accurately in real-time, which requires an accurate distance estimation framework. In this paper, we propose a novel Recurrent Neural Network (RNN) based architecture for accurately estimating the distance by the user using the smartphone IMU sensors. The proposed method combines signal processing techniques with RNNs leveraging the Long Short Term Memory (LSTM) layers to estimate the distance traveled by a person in real-time. Experimental analysis performed in different scenarios shows that the proposed method provides an average accuracy of 97.98% and 95.98% when tested over a travel distance of 100 m and 400 m, respectively.